An Evolving Resource
Industry
Same Datasets, New Perspectives
ABOUT THE AUTHOR
Margarita Pavlova, an experienced geophysicist with 14 years working in the resource sector provides her insight into tran-sitioning between the Oil and Gas and the Mining Industries where she has demonstrated there is ample opportunity for transfer of workflows, technology and knowledge.
CONTACT DETAILS [email protected] https: linkedin.com/in/ margarita-pavlova-16966616
MARGARITA PAVLOVA
Principal Geophysicist at BHPIt all started back in 2000 when I was mesmerizingly watching the Sydney Summer Olympics back in my hometown
Kaliningrad, Russia. The appeal of the land down under stuck with me and in 2001 I packed one suitcase and came to Australia on a student visa.
During my Marine Science Degree at Sydney University I studied a variety of Geology and Geophysics subjects and I chose to do a summer internship and Honours in the
Geophysics Laboratory run by Professor Ian Mason. My Honours project on borehole radar application to platinum reefs
introduced me to Mining Geophysics.
After Honours I was planning to continue with my PhD, but instead I was lured into Industry by Schlumberger in Adelaide. 2006 was a great time to start a career in the Oil and Gas industry and I travelled extensively with Schlumberger and learnt about the O&G business, geoscience and of course Petrel software – the flagship of Schlumberger Information Solutions. I taught numerous Petrel courses in Adelaide and around Australia and worked in Petrel and GeoFrame support. I have also consulted in data management, geomodelling and seismic interpretation projects.
Adelaide turned out to be much better than expected for a small town and I enjoyed ev-erything it had to offer including access to amazing wineries, beautiful beaches, great food and yes, cool nightlife. There I met my husband and in 2009 we moved to the UK for one year where I continued working for Schlumberger, doing on-site support to Nexen, BP, ENI, Anadarko and others.
They were very exciting times when I joined Origin Energy in 2010 and moved to Brisbane. The CSG industry was booming and all companies were in a race to construct LNG processing plants and to build teams to support them. After 3 years in the CSG business I made a
2
https://qurex.com.au/transition to the conventional O&G team. Throughout my work at Origin I had the great opportunity to be mentored by Randall Tay-lor – a superb geophysicist and a great guy. The majority of my time at Origin I worked in the Geophysics team, helping different as-sets with seismic interpretation, Quantitative Interpretation (QI) and occasionally geomod-elling.
After Origin Energy sold their convention-al assets to Beach Energy, I decided that I wished to stay in the Sunshine State and I made a transition to BHP - Minerals Australia and joined the BHP Coal Geophysics team. In 2017 BHP Coal made a business case to ac-quire 3D seismic over all of their Life of Asset mine areas and they embarked on an exten-sive seismic program. Seismic was not new to BHP Coal as BHP underground mines had utilised 3D seismic long before 2017. The new approach was that BHP had started acquir-ing high-density 3D seismic over their open cut mines. As a person who has worked with seismic throughout my career in both CSG and conventional settings I have a few things to “bring to the table”. Even though the main usage of the BHP Coal seismic data is to pro-vide mines with structural information, faults in particular, there is also big QI potential of seismic which I am exploring. A lot of QI tech-niques used by the O&G Industry are applica-ble to shallow coal surveys. Examples of this are seismic inversion and seismic attribute analysis including spectral decomposition. It is important for the mines to know elas-tic properties of coal seams’ overburden and interburden as well as coal properties and thickness of coal seams. Knowing the overburden properties of rocks such as rock velocity, density and lithology helps Geotech-nical and Drill and Blast Departments in their workflows and planning. UCS (Unconfined Compressive Strength) estimation is required for geotechnical modelling and for planning the type and quantity of explosives required in different areas to blast overburden and interburden effectively. Wireline sonic veloci-ty is correlated to UCS and used to construct 3D UCS models. It is possible to use velocity inverted from seismic or Surface Wave
Analy-sis velocity as a trend to guide interpolation of sonic data. Coal properties relate to coal qual-ity which includes a series of parameters with ash percentage being the most meaningful in relation to the seismic method as it has a strong link to density.
The products from Hampson and Russell’s post-stack and pre-stack simultaneous in-versions can be used to understand lithol-ogy distribution in the roof and floor of coal seams. However, these products are not useful for coal property definition due to two reasons: most of the target coal seams are 1-5 m thick and are below seismic tuning thick-ness; coal quality is mainly associated with minor density, acoustic impedance and Vp/ Vs changes, making it very hard to separate clean versus inferior coal.
Figure 1: Crossplot between density and
compressional velocity (Vp), colour-coded by facies in RokDoc. Sand and shale can be differentiated using velocity, while clean and inferior coal fit within a small velocity range and can be better separated using density. More advanced seismic inversion algorithms such as Ikon Science Joint Impedance and Facies Inversion (JiFi) and CGG Geostatistical Inversion are more suited to investigation below tuning thickness.
As I was exploring the seismic methodologies, which allow for
investigation of coal thickness and coal quality parameters including ash %, I was surprised that the extensive wireline log database was not utilised in coal quality modelling. The only input used for coal quality models was proximate analysis and washability undertaken on core samples. A big difference from O&G to Coal mining is the number of boreholes. Some areas are very densely drilled (less than 100 m spacing, particularly in structurally complex areas). Most wells drilled to only 150-250 m depth - chip boreholes are inexpensive to drill. They comprise 90-95 % of borehole database with the remainder being core holes. Most of the holes have a full suite of wireline logs, such as density, gamma ray and sonic. Core holes are more expensive to drill and proximate analysis is also more costly and time consuming, so the coal quality dataset is sparse.
It made perfect sense for me to first apply my prior Machine Learning (ML)
experience to predict coal quality parameters from wireline logs at BHP. I initiated a project with the Global Technology team to utilise the most modern ML techniques to solve the problem. We were achieving good metrics on the validation dataset, but there was overall some mistrust on how the ML predicts coal quality parameters using chip hole data. I implemented the Leave One Well Out (LOWO) test (commonly used to estimate velocity / depth uncertainty in O&G) to evaluate the usability of predictions in the chip holes. The LOWO results showed that adding ML predictions from chip boreholes to lab data reduces gridding error and ultimately reduces uncertainty. It also helped to compare performance of different ML algorithms and parameters. After initial success we are now exploring other Machine Learning applications to streamline the interpretation of our large borehole dataset.
Figure 2: Seismic section showing most probable facies from Geostatistical Inversion of shallow coal targets. We are able to match coal thickness despite it being below seismic resolution. The facies log of the blind well (not used in the inversion and marked by a star) matches the inverted section.
4
https://qurex.com.au/Figure 3: Percentages of raw ash from lab data (large circles) and ML predictions (squares) for
coal seams from two different mines. Predictions help to “fill the gaps” in the sparse lab dataset and visualise the lateral trend in the ash distribution.
To help me achieve some of my seismic interpretation and data wrangling goals I have been heavily relying on the O&G software applications such as Petrel, Hampson & Russell, RokDoc and PaleoScan. They have been our essential workhorses to understand and interpret shallow coal seismic and
better characterise our resources. What makes my job exciting is that at BHP Coal we have all the benefits of utilising new, sophisticated O&G technology like Paleoscan’s automatic fault picking and Hampson and Russell’s automatic synthetics correlation.
I often share ideas with BHP Petroleum geophysicists, who are recognised experts in geophysical technology advances and workflows. My team’s interactions are not limited to “soft rock” and occasionally we collaborate with Potash, Western Australia Iron Ore and Nickel businesses.
As somebody who made the transition from O&G to the Mining Industry I can say that there are numerous skills and fundamental knowledge that can be transferred from one
industry to another. People at BHP come from various technical backgrounds and I believe that BHP’s commitment to diversity has been instrumental to the company’s success. I find that the Mining Industry has a lot of opportunities to offer and I’m enjoying the high paced, open minded and welcoming nature of the Industry.
